Legora Says Legal AI Is Moving From Task Assistance to Matter-Level Agents

Max Junestrand describes Legora’s current product shift as a move from helping lawyers complete individual tasks to agents that can act across a legal matter. The earlier framing was augmentation: a lawyer asks for help in the context of a specific document, review, or drafting task. The newer framing depends on stronger models, Legora’s agent harness for legal work, access to enterprise content, and the trust required to operate inside large law firms and legal departments.

That change matters because the work is no longer just a chat or document-assistance experience. Junestrand says Legora can access documents, emails, matter context, tools, and customer-specific workflows inside large enterprises. In his example, a law firm partner opening an inbox with hundreds of messages should not first have to sort through the mess. Legora should have already looked across the context of the active matters and done useful work on the lawyer’s behalf.

The example he gives is an M&A transaction. In due diligence, a buyer-side legal team may receive an entire data room, often unstructured. Junestrand says a Legora agent can manipulate the file tree: the lawyer can provide a template folder structure and ask the agent to organize the room. The agent can then be asked to apply diligence questions for a specific type of company and identify missing content. Some of these jobs run for 20 or 30 minutes, which changes the user interaction. The lawyer is no longer working with the system only in real time. They are giving broader instructions and letting agents work in parallel, in the same broad pattern Junestrand associates with tools like Cursor or Claude Code.

The product bottleneck, in that world, becomes evaluation. Junestrand says Legora is moving toward “end-to-end work products,” not just individual task outputs. That creates a harder question than whether a generated paragraph is acceptable: the company has to assess whether a multi-step legal workstream was completed correctly enough to be useful in a professional setting.

Junestrand also places legal work behind software development on the adoption curve. Code, in his view, will remain at the frontier because it is easier for models: it is more binary, has a simpler context problem, and models have been trained heavily on it. Legora watches coding tools as an indicator of where legal agents may go next, with roughly a six-month lag.

Legora’s growth did not come from narrowing to a single point solution, according to Junestrand. In October 2024, when the company went into general availability, it had about 30 people and three core product bets: an agent and assistant, a “tabular review” product, and a Microsoft Word add-in. The company wrote a three-page product manifesto saying it would become the best in all three.

That was not the obvious strategy if judged by immediate market position. Junestrand says one company focused only on the Word add-in and another focused only on tabular review. The tabular-review competitor was doing almost $50 million in ARR while Legora was doing about $1 million. In other words, a competitor with roughly 50 times Legora’s revenue was concentrating on one wedge while Legora was trying to bundle three.

Junestrand’s account is that Legora held to a longer product horizon in a fast-moving market. If the company could become best in chat, best in tabular review, and best in Word, then bundling those pieces could beat a more opportunistic, single-feature strategy. He says Legora has since “enormously surpassed” the focused competitor. He also says, in the same passage, that Legora has “churned so many other clients”; he does not clarify the phrase further, so the stronger point is the one he makes explicitly: even in a short-cycle AI market, founders need a view of how the product world should look years out.

Junestrand says Paul Graham gave the company “fun advice” at the YC office: write a science-fiction novel about what a lawyer 10 years from now will do and how they will work. “At the end of the day,” Junestrand says, “it’s that future we are building for.”

That line captures the operating tension in Junestrand’s view of AI startups. Legora was moving quickly enough to compete in live enterprise sales, but the product strategy was organized around a future lawyer’s workflow rather than whatever feature was easiest to sell that quarter.

Gustaf Alströmer raises the question many AI founders now face: what happens if OpenAI or Anthropic builds the same thing? He compares it to an older startup anxiety, “What if Google does this?” In his telling, that was a serious question when Google looked capable of entering many adjacent categories, but over time founders became less intimidated by it because Google did not execute on many new internal product efforts.

Junestrand answers by reframing the competitive threat. The question is not only whether a foundation-model company will enter a market. It is what remains defensible as models become smarter, whether that improvement is linear or exponential. If models eventually solve every part of every task on the fly — writing the code, gathering the data, and deciding the workflow unaided — then, in Junestrand’s words, everyone may as well “go have a piña colada.” He does not think that is the likely end state.

Instead, he points to a different set of moats: inputs, outputs, proprietary data, workflow position, and user behavior. For Legora, that meant getting big fast. The value, in his account, is not just in calling a model. It is in being embedded inside the work, having access to the relevant enterprise content, shaping how legal professionals operate, and building around the workflows in which work is requested, completed, reviewed, and repeated.

He compares the dynamic to databases and infrastructure in the AWS era, citing MongoDB as an example of a company that had to position itself against a much larger platform and build what was not natural for that platform to build. The analogy is not developed into a detailed competitive map, but the implication is clear: a startup has to decide which product, workflow, and data advantages a platform provider is unlikely to replicate with the same focus.

Junestrand says Legora did not begin with a great product. What it had, in his telling, was intensity, clarity about the market, and a sales posture that felt unusual in legal technology.

After entering YC, the company had roughly 10 people. The engineering team went to San Francisco and lived together in an Airbnb. Junestrand describes the period as a “work camp”: engineers shipping, sales calls running between 1 a.m. and 10 a.m., and the team eating bad food while grinding through the batch. He says Legora entered YC with both confidence and imposter syndrome. They expected to meet companies led by PhDs from MIT and Google that already had revenue and knew exactly what they were doing. Instead, they found many companies still searching, while Legora already knew what it was building and had among the highest revenue in the batch.

That clarity changed how the team used YC. Junestrand moved back to Sweden, “bunkered” in a conference room, and sold. He recalls running around Stockholm with a briefcase and pitching chief innovation officers, knowledge managers, and legal partners with an energy they had not seen from legal-tech vendors. His pitch, as he paraphrases it, was blunt: this is the future, Legora will help you win, and the largest firm in the Nordics already works with us, so not joining means falling behind.

The product, by his own account, was “frankly not that great.” But buyers wanted to work with the team. Junestrand attributes part of that to excitement. Legal-technology sales was associated with tired pitches and bland positioning. Legora presented itself as urgent and ambitious in a category not used to that tone.

The team then performed what Junestrand calls a tactical “hot swap.” The engineers and product team returned from the United States to take care of customers. Junestrand went to YC to fundraise.

For a first-time founder with little network, Junestrand says one underappreciated part of YC is the investor reach and signaling value. Investor inbound builds toward Demo Day, and the practical move is to compress the meetings. He describes lining up roughly 80 investor meetings in a week and “crunching it out.”

The process was not simply automatic. Junestrand says Legora’s practice pitch at YC was bad; Alströmer adds that they were tired and unprepared. But Junestrand says he performs when the stakes are real. In the live fundraising environment, he says, the company “smashed it.”

His Benchmark story carries the tension of being an ambitious European company pitching elite Silicon Valley capital. Junestrand recalls a 30-minute pitch with Peter Fenton and Chetan Puttagunta, who later joined Legora’s board. After the meeting, Junestrand says, Fenton told Puttagunta: “The guy is perfect. The only problem is that he’s from fucking Sweden.” Junestrand’s response on stage is that he does not think that will be a problem anymore.

Alströmer adds a founder-level observation about fundraising psychology. As founders accumulate investor rejections, they can start treating the no’s as evidence that the investors are right and the company is wrong. Each meeting makes them a little less confident, and investors can detect that loss of conviction. To get others to believe the company will succeed, the founder has to continue believing it.

Junestrand says he can create that confidence “on the go.” But he also links confidence to a longer commitment. Some companies work and some do not; Legora’s founders decided that this company would be their life’s work. Once he made that decision personally, he says, the amount he cared and the scale of his ambition increased.

Max Junestrand does not describe Legora’s long-term ambition as becoming only the dominant legal AI vendor. He says the company wants, one day, to remove the word “legal” from “legal tech Legora.”

His model for company-building is breadth. He calls Google one of the best companies in the world because its search and advertising business created the opportunity to build under the Alphabet umbrella in areas like self-driving cars. He also cites Facebook’s attempt to build Meta as a “cool idea,” even while saying the company has been “fumbling the ball a bit.” The point is not that Legora is already equivalent to those companies; it is that Junestrand wants a core business strong enough to let the company attack much larger adjacent opportunities.

That ambition is also explicitly European. Junestrand argues that Europe’s largest technology company being SAP is not good enough. He sees the current AI moment as a rare opening in which leverage can shift. He applies the same logic to law firms: if a firm is ranked number 150 in the United States, AI may be its ticket to move toward the top 10. The broader claim is that technology is democratizing access to tools and talent, and the missing ingredient is ambition.

Legora’s hiring strategy appears tied to that view. Junestrand says the company calculated that 15% of its engineering and product organization consists of ex-YC founders. Product departments are often run by former CEOs. He describes Legora as a company run by founders throughout the organization, with “founder mode energy” distributed beyond the executive team.

That founder density is part of how he explains the company’s ability to scale quickly without losing intensity. The company had about 40 people a year before the talk and is now close to 500, spread across San Francisco, Chicago, Texas, New York, London, Stockholm, Germany, India, and Australia. Junestrand says the company is going global because lawyers work in broadly similar ways around the world.

He frames the current stage as “base camp,” not the summit. Legora has proven one part of the journey to itself, but he says nobody inside the company is content.

Asked how he thought about his options while in school, Junestrand says he tried to do as much as possible. He studied some computer science, studied some business, worked at McKinsey, and worked at two YC startups. He says he spent only one week at Depict before its founder, Anton, tried to hire him; Junestrand declined because he wanted to finish his degree.

He does not present legal AI as a carefully selected market from a spreadsheet. “Did we pick law or did law pick me?” he says. “I think law picked me.” Once that happened, the team “decided to run like hell” and did not seriously consider doing something else.

The risk calculation was initially manageable. Junestrand had a full-time McKinsey offer “in the back of the pocket,” so working on Legora over the summer did not feel especially risky. YC acceptance changed the decision. That is when he called McKinsey and said he was not coming back.

He says this pattern became common at Legora. Ten to fifteen people at the company had offers from McKinsey in particular. Jake, Legora’s CTO and a previous YC founder, had deferred his McKinsey offer for six years after an internship. Eventually, when the firm called again to ask whether he was coming back, he said no.

The YC timing was unusually favorable. Legora began working in summer 2023 and applied through an early AI application process for the winter batch. Alströmer says that batch was not necessarily the peak AI batch but was the first “best AI batch.” Junestrand says YC’s early acceptance in August bought the company time to prepare before the batch began. When Alströmer called to ask whether he would move to San Francisco, Junestrand says he answered yes because that was the correct answer — though he remains in Stockholm.

Legora’s Jude Law campaign began, in Junestrand’s telling, from a simple observation: marketing for legal technology is usually “the most boring, the most bland.” The team had an “AI-powered law” slogan and, after at least one bottle of wine in the office, joked about getting Jude Law to say it.

The first agency response was that Jude Law was impossible to get. Junestrand also notes a broader Hollywood context: many actors and screenwriters are anti-AI because AI is beginning to affect scripts, color, and cinematography. For an actor to stand behind an AI company would be a meaningful public stance.

Legora chased Law for about six months. He initially said no. The company then tried to persuade him through the product and customer evidence rather than the slogan. Junestrand says they showed him testimonials from lawyers, including quotes collected in a company Slack channel called “customer love.” One example he gives is a lawyer saying they used Legora to review a thousand agreements in one day and got home to see their family before the weekend.

That changed the conversation. According to Junestrand, Law said he could get behind the product, but he wanted to remain Jude Law rather than become a corporate mascot. The line became “there’s a new face of Law.” Law brought his own screenwriter and cinematographer; Junestrand says the campaign ended up involving a Saturday Night Live scriptwriter and the cinematographer of Oppenheimer.

The result was not just a polished ad. Alströmer says he saw the video and then saw it everywhere when he landed in Stockholm. Junestrand says the campaign created “17 touchpoints” and even produced a lead through someone’s mother recognizing “this Legora thing” because of Jude Law.

The anecdote is funny, but it also fits Legora’s broader positioning. Junestrand is explicit that legal technology had trained buyers to expect dullness. Legora wanted the category to feel more alive and culturally visible — the same posture it used in early sales, translated into marketing.